Multi-scale Image Analysis of Satellite Data Using Perceptual Grouping

A meaningful image segmentation groups the pixels into disjoint regions that consist of uniform components. Facing absence of contextual knowledge, the only alternative which can enrich our knowledge concerning the significance of our segmented groups is the creation of a hierarchy guided by the knowledge which emerges from the superficial and deep image structure [1]. In the present work, we deal with information which can be retrieved from the superficial structure of the image and we study hierarchical methods based on a region-based approach, wherein the initial pixel grouping is guided by the principles of the watershed analysis [2]. Hierarchical feature representation through multi-scale segmentation offers new possibilities in object-oriented and multi-scale image analysis of satellite images [3, 4]. Our objective, in this work, is to create a hierarchy among the gradient watersheds which preserves the topology of the initial watershed lines and extracts homogeneous objects of a larger scale. The proposed method segments an image into regions, which are then merged using an innovative saliencydriven perceptual organization approach. Moreover, we propose a criterion which allows automatically extracting the best segmentation level from the obtained hierarchical levels.